Traditional dietary methods are constrained by digestion biases, prey misidentification, and short temporal resolution. Stable isotope analysis (SIA) overcomes these limitations by providing a time-integrated record of nutritional dependence via δ¹³C (resource origin) and δ¹⁵N (trophic level), enabling isotopic niche modelling of resource use across habitats and disturbance gradients. Dorsal muscle remains the standard tissue for fish SIA, but its lethal collection is problematic in no-take zones and for threatened species; sublethal alternatives such as fin clips require validation before adoption. Parma polylepis (Banded Scalyfin), a site-attached, habitat-generalist damselfish ubiquitous across Norfolk Island's coral and rocky reefs, is well suited to detecting localised trophic variation, yet Norfolk Island's trophic ecology remains unstudied, and Kingston Lagoon's no-take status precludes lethal sampling there. We aimed to (1) characterise the isotopic niche of P. polylepis across inshore coral, fringing coral, and rocky reef habitats, and (2) evaluate dorsal fin clips as a sublethal proxy for muscle tissue. Fish were lethally sampled via handspear at five sites, with sublethal fin clip sampling at Kingston Lagoon and used for subsequent SIA and niche analysis. P. polylepis at Elephant Rock were isotopically distinct, with elevated δ¹⁵N, broader δ¹³C distributions, and negligible pairwise niche overlap (<2–7%) relative to all other sites. The remaining five sites occupied broadly similar isotopic space, with Anson Bay fish forming the most constrained niche, nested within the broader space (67–90% asymmetric overlap), and Cemetery Bay, Kingston Lagoon, and The Cord showing high mutual overlap (55–92%), indicating similar resource use. Fin and muscle δ¹³C and δ¹⁵N differed significantly (p < 0.001), and the fin–muscle δ¹⁵N relationship was strongly site-dependent (R²C = 0.83), precluding a universal correction factor. Fin clips offer a practical sublethal proxy where lethal sampling is restricted, provided site- and individual-level variation is accounted for with adequate sample sizes.